Functions are not so rare at all, and definitely not isolated, in sequence space of biopolymers

[Witchdoc]

Promiscuous enzymes could probably be used as an argument against design.

In humans, alcohol dehydrogenase catalyses the oxidation of methanol into formic acid, which causes blindness in humans.

A better designed alcohol dehydrogenase would not have this downside.

[Dan_Eastwood]

@CrisprCAS9

I don’t follow your question in the context of my previous comment. Sorry.

No worries, I could have included Giltil’s claim too, and I didn’t explain myself well.

@CrisprCAS9

@Giltil

Yes, but there are still immensely more bunch of ways not to do anything at all.

Define ‘immensely’. 5-10 orders of magnitude? Sure. The often suggested >100 orders of magnitude? Certainly not. Having a one in a billion chance of getting a functional protein from de novo translation is entirely sufficient.

@Dan_Eastwood

That has to be getting close to saying that function doesn’t exist at all. What do we get if we multiply the number of known functions by 100 orders of magnitude?

Now let me add to this …

Since each of the 20 amino acids is chemically distinct and each can, in principle, occur at any position in a protein chain, there are 20 × 20 × 20 × 20 = 160,000 different possible polypeptide chains four amino acids long, or 20^n different possible polypeptide chains n amino acids long.
[source]

So if function is immensely rare, by >100 order of magnitude, then we shouldn’t expect to find much of any function in chains less than 100 peptides long. We do find function in shorter chains, therefore function cannot be so incredibly rare.

Said another way, if function really were so incredibly rare, then some of the functions we already know should not exist at all.

[CrisprCAS9]

Ah, I think then you are asking a question of Giltil, and not me.

I think the notion of rare functionality is obviously false. As I said, it might be true that only 1 in a million (or even 1 in a billion) random strings has relevant functionality, but certainly not 1 in 10^30 (or 10^170 as Axe suggests). This is not a guess on my part, it has been tested experimentally.

[Michael_Okoko]

@CrisprCAS9

This is not a guess on my part, it has been tested experimentally.

References please.

[CrisprCAS9]

Screening of peptides recognizing simple polycyclic aromatic hydrocarbons

Or just search ‘phage display library’ in your favorite database.

[Giltil (Gilbert Thill)]

Keep in mind that for a protein the ability to bind something is quite a simple function and as such, it is not surprising that proteins with this ability are not so rare. But things are very different for proteins carrying more complex functions such as, for examples, DNA polymerases or ion channels.

[Giltil (Gilbert Thill)]

@Dan_Eastwood

Said another way, if function really were so incredibly rare, then some of the functions we already know should not exist at all.

Not all functions are rare in sequence space, only complex ones.

[T_aquaticus]

@Giltil

But things are very different for proteins carrying more complex functions such as, for examples, DNA polymerases or ion channels.

Urf13 evolved through recombination of mitochondrial genes, and it has ligand-gated pore forming activity.

URF13, a ligand-gated, pore-forming receptor for T-toxin in the inner...

[CrisprCAS9]

@Giltil

Keep in mind that for a protein the ability to bind something is quite a simple function and as such, it is not surprising that proteins with this ability are not so rare.

I agree…

@Giltil

But things are very different for proteins carrying more complex functions such as, for examples, DNA polymerases or ion channels.

How do you think those things work? Binding to the strand (easy), binding to the nucleotide (easy), conformation change (easier). That’s basically all biochemistry: Binding affinities and conformation changes.

[Rumraket (Mikkel R.)]

@Giltil

But things are very different for proteins carrying more complex functions such as, for examples, DNA polymerases or ion channels.

T-urf13 is an ion channel and evolved during selective breeding of maize.

On the evolution of Irreducible Complexity

The novel antimicrobial peptide discovered and described in Knopp et al 2019 is also an ion channel:
https://journals.asm.org/doi/full/10.1128/mBio.00837-19

Generally speaking transmembrane proteins are surprisingly easy to evolve, so much so that they are among the most likely de novo proteins.

De novo emergence of adaptive membrane proteins from thymine-rich genomic sequences

Another intriguing de novo gene evolution paper was recently published in Nature Communications: De novo emergence of adaptive membrane proteins from thymine-rich genomic sequences Some excerpts: Mutations that cause changes to the sequence or expression of established genes are typically constrained by preexisting selected effects—the specific physiological processes mediated by the gene products that are maintained by natural selection20. In contrast, emerging proto-genes are expected to mostly lack such constraints because they do not have selected effects. This would leave them more readily accessible to evolutionary changes that have the potential to increase fitness (adaptive changes)3,4. We reasoned that this initial potential for adaptive changes would give way as proto-genes mature and the adaptive changes engender novel selected effects, in turn increasing constraints and reducing the possibility of future change. This reasoning is akin to Sartre’s “existence precedes ess…

The very weak sequence constraints on transmembrane domains has also been shown phylogenetically:
https://academic.oup.com/mbe/article/33/11/2874/2272007

Transmembrane domains are so easy to evolve because they’re essentially just repeat-proteins consisting of a single, simple structural element, such as a beta-hairpin. These will naturally tend to oligomerize, their hydrophobic exterior will have an intrinsic affinity for the hydrophobic interior of the membrane bilayer, and they form so-called “barrel” (essentially just tube-shaped structures).

(http://membranproteine.net/Structure%20gallery%201.html)

There is a lot of literature on the evolution of transmembrane repeat proteins (such as beta-barrel structures).

Evolutionary pathways of repeat protein topology in bacterial outer membrane...

https://www.sciencedirect.com/science/article/pii/S0969212618302132

Etc. etc.

[Rumraket (Mikkel R.)]

@Giltil

Not all functions are rare in sequence space, only complex ones.

I should add that I don’t think you’re wrong to say that some functions are more, or even considerably more rare, than others. I don’t think it would be sensible to extend the hyperabundance of some specific function in sequence space, to basically all protein functions. That would be an unwarranted extrapolation, and also doesn’t make biochemical sense.

But some functions being much more rare than others does not in and of itself imply they are too rare to evolve. It might imply they’re unlikely to evolve by de novo evolution from non-coding DNA, but there are other mechanisms available to evolution for novel gene evolution than that. I wrote a post about that here:

@Rumraket

@colewd

I would argue if you traced its origin back to the first enzyme of a similar type the design inference becomes a viable explanation for the origin of catalytic activity in an enzyme containing hundreds of amino acids.

There doesn’t appear to be any need of that hypothesis.

While I think Axe’s estimate of frequency of sequences that adopt the structure and function of beta-lactamases is an extremely skewed underestimate (in part because of numerous shortcomings of his experimental approach), there does appear to be some evidence that comparatively speaking enzyme sequences are considerably more rare in protein sequence space, than things like ligand/molecule binding proteins, transmembrane transporters, or transcriptional regulators and the like.
In this respect I don’t think it would be wrong to say that you’d expect de novo protein evolution (from non-coding DNA) to yield enzymes much more rarely than, for example, proteins with structural or regulatory roles.

However, I think it is a mistake to take this to imply that de novo enzyme evolution is so rare as to be essentially impossible. One problem is there are numerous ways to get a de novo enzyme sequence than just the spontaneous emergence of an ORF in some stretch of non-coding DNA. And while on the subject of non-coding DNA, even this is more likely to yield something functional, than the sort of spontaneous polymerization “tornardo-in-a-junkyard” picture one gets from reading IDcreationist material. We have to be mindful that a lot of non-coding DNA is actually previously functional, protein coding DNA that has been carried along by selfishly expanding transposable elements. That means a lot of non-coding DNA is actually quite near in sequence space to DNA encoding significant secondary structural elements of proteins. Much nearer than just completely random.

That said, I think a more holistic view of protein evolution implies that enzyme functions are most likely to emerge in already existing, stably folding proteins, or (less often) from gene fusion events where pieces of already existing genes are copy-pasted as insertion vents, resulting in shuffled and recombined protein coding gene fragments. Folding proteins are largely modular structures consisting of secondary structural elements, which can in principle be moved around and combined into new structures.

This makes much more sense to me, because pieces of already existing proteins are of course much nearer in protein sequence space to “something potentially functional”, than some arbitrarily picked non-coding DNA sequence that has been blindly accumulating mutations for eons. So when it comes to explaining the ultimately origin of novel catalytic functions, I would rank (from more to less likely) the order of likelihood of explanations like this:

1. Emergence of new catalytic function in an already existing catalytic scaffold (divergence of duplicate enzymes).
2. Emergence of new catalytic function in a non-catalytic scaffold (structural/ligand binding protein becomes an enzyme).
3. Domain or subdomain sized fragments(exon shuffling, insertions, or just gene-gene-fusion) combine to yield a new protein with a catalytic function. (see this and this and this)
4. Oligomerization of smaller protein fragments result in a new protein (repeat-proteins).
5. De novo gain of an open reading frame that codes for a functional enzyme sequence, from non-coding DNA.

I think the vast majority of enzyme function gains in the history of life owe to 1 and 2, with the diversity of all known chemical reactions catalyzed by enzymes reducing to a smaller set of enzyme superfamilies. There’s a nice article showing exactly this here:

journals.plos.org

Exploring the Evolution of Novel Enzyme Functions within Structurally Defined…

Author Summary Enzymes, as biological catalysts, are crucial to life. Understanding how enzymes have evolved to perform the wide variety of reactions found across all kingdoms of life is fundamental to a broad range of biological studies, especially…

Now these families in turn have to have originated somehow, and here I think 3 and 4 provide the best, most likely explanations for the reasons already stated. Pieces of already existing, stably folding proteins are much closer in protein sequence space to something potentially functional, so their shuffling/recombination and fusion into larger structures constitutes a much more efficient, and biased-towards-functional-and-folding “search process” than spontaneous gain of an enzyme fold from non-coding DNA does. And there’s a lot of good evidence that stably folding tertiary structures can be gained by repeating duplications (oligomerization) of smaller fragments.

This is, again, what I consider a more holistic view of protein evolution over the history of life on Earth. It’s the sort of picture I’ve come to hold after 13 years of reading the literature, and trying to follow the evidence and arguments from different camps. To sum up, I think you put too much trust in Axe’s work, and when considering evolution I don’t think you really appreciate the many different “search strategies” are available to evolution besides the sort of “randomly hook amino acids together into longer chains” pop-out-of-a-soup-of-amino-acids view.

[Rumraket (Mikkel R.)]

@CrisprCAS9

How do you think those things work? Binding to the strand (easy), binding to the nucleotide (easy), conformation change (easier). That’s basically all biochemistry: Binding affinities and conformation changes.

I think each of those are likely enough, but their conjunction is probably considerably more unlikely in the same structure. If each function on it’s own has a probability of 10^-9, back of the envelope estimate you get an overall probability of 10^-27. That would make it quite hard to evolve from non-coding DNA, which is why things like shuffling of fragments, recombination among sub-domains structures, and gene-fusion are very important in evolution, which is why work like this is very relevant and important to this question:

https://academic.oup.com/mbe/article/38/6/2191/6120801

[CrisprCAS9]

@Rumraket

which is why things like shuffling of fragments, recombination among sub-domains structures, and gene-fusion are very important in evolution

Oh sure, absolutely. Then again, each of those individual functions (potentially) has intrinsic utility to an organism, especially an organism that didn’t have a system with all three together. Which helps explain where you got the material for those recombinations.

[Mercer (John Mercer)]

@Giltil

Keep in mind that for a protein the ability to bind something is quite a simple function and as such, it is not surprising that proteins with this ability are not so rare.

Only if you keep in mind that binding is the essence of catalysis, which is very basic biochemistry. That’s why we can immunize mice with a substrate, with the immune system selecting ONLY for binding, and end up with catalytic antibodies.

@Giltil

Not all functions are rare in sequence space, only complex ones.

Myosins are among the best-studied and most complex proteins known.

Why is human beta-cardiac myosin so polymorphic, while the obviously less-complex actin is not at all, Gil?

This shows that @gpuccio’s notion that sequence conservation represents functional complexity is absurd.

[RonSewell (RonS)]

@Giltil

Keep in mind that for a protein the ability to bind something is quite a simple function and as such, it is not surprising that proteins with this ability are not so rare.

You are on the path to enlightenment here.

Binding is pervasive in biochemistry. Bonds - ionic bonds, covalent bonds, hydrogen bonds, van der Waals interactions - binding electrons, binding simple diatomic molecules, and binding gigantic biomolecules. We have tracked while SARS-CoV-2 synergistically enhanced its binding to ace2 receptors through multiple mutations; with lethal effect for contagion and virulence. The same mutations have independently occurred in multiple strains, so the function in sequence space could not be so rare. Influenza virus normally uses hemagglutinin for cell entry and neuraminidase for egress, but the virus has been observed to mutate neuraminidase to resourcefully achieve receptor binding in place of hemagglutinin.

@Giltil

But things are very different for proteins carrying more complex functions such as, for examples, DNA polymerases or ion channels.

Not so different. DNA polymerases and ligand-gated ion channel channels involve binding.

A great deal of our body’s homeostasis involves binding. Binding by one molecule can change the shape of the bound molecule, causing the release or binding of yet another molecule. Extensive feedback loops function by complex cascades of binding. Nature is always fiddling with the electrostatic shapes of biomolecules, and while complexity is not a goal it is often a result. Indeed, it isn’t at all surprising that the proteins which perform the functions which sustain life are not so rare.

[Giltil (Gilbert Thill)]

@RonSewell

Not so different. DNA polymerases and ligand-gated ion channel channels involve binding.

The bindings involve in complex functional proteins such as DNA polymerases or ligand-gated ion channels are not mere unspecified bindings, but highly fine tuned, integrated, coordinated and regulated bindings which can only be established within large proteins exhibiting high functional information.

Why Are Enzymes So Big? – Creative Enzymes Blog

@RonSewell

Nature is always fiddling with the electrostatic shapes of biomolecules, and while complexity is not a goal it is often a result. Indeed, it isn’t at all surprising that the proteins which perform the functions which sustain life are not so rare

Your second sentence doesn’t follow from your first one, quite the contrary.

[Witchdoc]

@Giltil

The bindings involve in complex functional proteins such as DNA polymerases or ligand-gated ion channels are not mere unspecified bindings, but highly fine tuned, integrated, coordinated and regulated bindings which can only be established within large proteins exhibiting high functional information.

Why Are Enzymes So Big? – Creative Enzymes Blog

@RonSewell

Nature is always fiddling with the electrostatic shapes of biomolecules, and while complexity is not a goal it is often a result. Indeed, it isn’t at all surprising that the proteins which perform the functions which sustain life are not so rare

Your second sentence doesn’t follow from your first one, quite the contrary.

But that second sentence is well supported, if you look at the scientific literature.

A paper that found that 8000 mutants of DNA polymerase MOTIF A ALONE (that is, only the 13 amino acid functional site) were functional, most of which had activity similar to the wild type; They found only ONE of the 13 amino acid was required to stay the same.

There would be so many more functional mutants if they mutated other parts of the enzyme.

The enzyme active site, is probably the part that can vary LEAST.

https://www.researchgate.net/publication/248529063_Conservation_and_mutability_in_molecular_evolution

For example, if we extrapolated 8000 functional sequences from varying the core 13 AAs to the rest of 130 AA protein, then there would be 8000^10 functional variants.

There are likely more than 8000^10 (ie, 10^39) 130AA functional variants of DNA polymerase A. And of course, more if you allow the number of AAs vary.

[RonSewell (RonS)]

@Witchdoc

A paper that found that 8000 mutants of DNA polymerase MOTIF A ALONE (that is, only the 13 amino acid functional site) were functional, most of which had activity similar to the wild type; They found only ONE of the 13 amino acid was required to stay the same.

The Patel and Loeb source paper here.

@Giltil

The bindings involve in complex functional proteins such as DNA polymerases or ligand-gated ion channels are not mere unspecified bindings, but highly fine tuned, integrated, coordinated and regulated bindings which can only be established within large proteins exhibiting high functional information.

It would appear that a high degree of optimization and conservation does not infer brittleness of functionality. See also:

Protein tolerance to random amino acid change

Do you consider the spike mutations of the SARS-CoV-2 delta variant to be a “mere unspecified binding”, or a “highly fine tuned, integrated, coordinated and regulated bindings which can only be established within large proteins exhibiting high functional information.”?

[Mercer (John Mercer)]

@Giltil

The bindings involve in complex functional proteins such as DNA polymerases or ligand-gated ion channels are not mere unspecified bindings, but highly fine tuned, integrated, coordinated and regulated bindings which can only be established within large proteins exhibiting high functional information.

If it’s that fine tuned, how was it that my colleagues and I could change the specificity of two of the most complex proteins known (myosins) without a significant effect on their normal functions, merely by changing a single amino-acid residue in the active site to one not found in nature?

If you are correct, we could not possibly have succeeded. Link to 5 papers below:

17951722,16102537,14766983,11853671,10531338 - Search Results - PubMed

[Faizal_Ali]

@Giltil

Yes, but there are still immensely more bunch of ways not to do anything at all.

Just as there are immensely more ways to lose a lottery than to win it.

Therefore, lotteries are never won.

Do you agree with that conclusion?